www.delorie.com/pcb/inkjet/   search  
Inkjet Photomask Experiments

Click on any thumbnail to see the full-size picture

This is the output of the bitmap-to-pcb converter. It keeps all the pixels near the edges, but removes half of the ones in the "field". I've modified my gutenprint config file to print at maximum ink density, which would be too much ink if I didn't remove half the pixels. By leaving the edge pixels in and manually dithering the field, I get a clean edge without too much ink. The bitmap resolution is 2880 DPI, which matches my printer.
This is a sample of 5/5 rules - 5 mil lines with 5 mil space. Although the printer is mechanically incapable of aligning dots within better than 1 mil accuracy, it's good enough for my purposes. There is enough clear to ensure an exposed film, and enough dark between to ensure that the copper is etched.
Same, but 6/6 rules. Note that these photos were taken with a microscope, lit from underneath the film - so the speckles you see are spots where the ink isn't quite as solid as other places. Should be OK though, but we'll see. At least the speckles are away from the edges - they're due to the pixels we removed earlier.
A sample via. This is a 13.5 mil hole and a 13 mil wide annulus. The trace is 12 mil. The pad is for an 0603 part.
I tried a print without the dithering - i.e. full ink. Seems to have worked, but then again, this is SilkJet film.
Likewise, on the 5/5 traces.
Likewise, on the 4/4 traces. At this point, the printer's ability to accurately feed paper is starting to show - that's what causes the "comb" effect on the vertical stripes (the film is viewed so its feed was left-to-right). Note that this varies from print to print; the 5/5 and 6/6 lines on this particular print show similar combing, whereas the 5/5 in the photo above shows less.
Likewise, on the 3/3 traces. It will be interesting to see how that film and copper stand up through etching.
This is the resulting etch from the 4/4 traces. The photos are vertical stripes (relative to the printer) on top, horizontal below, and front-lit on the left and back-lit on the right. Front-lit shows that the traces are well-formed, but back-lit shows the shorts from the thin bits of unetched copper. Another couple of minutes in the tank probably would have made this near-perfect. Note: the board has been tinned with Liquid Tin; next time I'll leave it untinned incase the tinning process affects the quality of the edges. I wouldn't think so, but at these sizes, it's best to be sure.
Compare with the larger 12 mil traces elsewhere on the board. Despite the inkjet's ragged edges, the copper edges are well-formed. I give credit to the photomask. Note that the relative sizes of trace vs line, here and above, is equal - so no need to adjust the edges.
Some vias. The holes are 8 mil, but you can see that all the edges are well-formed and smooth. The vias really are round; the microscope software doesn't give me the right aspect ratio.
Example of a 5/5 etch. The design has a 5 mil line going between two 13 mil pads, with 5 mil spaces. The actual copper is a 3 mil line with 6.5 mil spaces, and 10.3 mil pads. So, I need to bloat the edges by about 0.8 mil to get it to etch correctly. The pads are for an 0.5mm pitch CSP chip.
This is the mask for the CSP (different run than the above photo) showing how well the photomask works out.
And last but not least, the finished board for the above.

So that I don't forget the GIMP settings:

claria density set to 4.0 in /usr/local/share/gutenprint/5.1.7/xml/printers.xml
quality: best
media: glossy photo
dpi: 2880
direction: unidirectional
interleave: standard
ink: six-color photo

  webmaster   donations   bookstore     delorie software   privacy  
  Copyright 2008   by DJ Delorie     Updated Jul 2008